Anthropic's New Claude Opus 5.5 Comes With a Safety Brake

Anthropic introduced Opus 5.5, which the poster calls "a sizeable upgrade" — and, per the post, it's the first model in which Anthropic says this: after CEO Dario Amodei argued the week before that AI progress should be paced so safety practices stay ahead of model capabilities, the release carries Anthropic's pacing message. That message describes safety work on two time horizons: practices for current models, including extensive alignment testing, pre-release evaluation by outside organizations such as METR and Frontier Design, and safeguards matched to each model's capabilities in high-risk areas like cybersecurity and biology; and preparation for future models, including tighter filtering of reinforcement-learning environments, improved alignment rewards, automated generation of diverse safety-training scenarios, and stronger security and monitoring such as interpretability-based monitoring meant to reduce reliance on auditing a model's chain-of-thought. Anthropic adds that models capable of fully automating AI research itself would require a higher safety standard still, and that public policy should play a larger role as AI becomes more capable. In the comments, one user wrote that OpenAI "countered by boosting Sol and Luna to 6," linking to an OpenAI page — a comment, not part of the article itself, and no timing is stated.